TESTING: test just special case boolean code

[alamb]

• part of Speed up binary kernels by XXX%, add BooleanBuffer::from_bitwise_binary #9022

I am trying to figure out why the boolean benchmarks are showing performance differences when I don't think they are making any changes

This is the based on the baseline

• TESTING: no code change, boolean kernel benchmark stability #9069

But with the special case path from

• WIP: special case bitwise ops when buffers are u64 aligned #8807

I am hoping that the slice24 benchmark does not show any changes here

[alamb]

run benchmark boolean_kernels

[alamb-ghbot]

🤖 ./gh_compare_arrow.sh gh_compare_arrow.sh Running
Linux aal-dev 6.14.0-1018-gcp #19~24.04.1-Ubuntu SMP Wed Sep 24 23:23:09 UTC 2025 x86_64 x86_64 x86_64 GNU/Linux
Comparing alamb/test_bin_op2 (42f7ce7) to 2d6fc51 diff
BENCH_NAME=boolean_kernels
BENCH_COMMAND=cargo bench --features=arrow,async,test_common,experimental --bench boolean_kernels
BENCH_FILTER=
BENCH_BRANCH_NAME=alamb_test_bin_op2
Results will be posted here when complete

[alamb-ghbot]

🤖: Benchmark completed

Details

group            alamb_test_bin_op2                     main
-----            ------------------                     ----
and              1.00    205.7±0.65ns        ? ?/sec    1.33    274.4±3.60ns        ? ?/sec
and_sliced_1     1.00   1090.1±2.98ns        ? ?/sec    1.00  1095.1±14.63ns        ? ?/sec
and_sliced_24    1.00    241.0±6.00ns        ? ?/sec    1.00    240.7±3.51ns        ? ?/sec
not              1.01    146.9±3.28ns        ? ?/sec    1.00    145.8±3.81ns        ? ?/sec
not_slice_24     1.00    194.8±3.12ns        ? ?/sec    1.00    195.8±5.06ns        ? ?/sec
not_sliced_1     1.00   622.0±13.64ns        ? ?/sec    1.00    619.0±3.08ns        ? ?/sec
or               1.00    195.2±1.67ns        ? ?/sec    1.30    254.6±4.96ns        ? ?/sec
or_sliced_1      1.00  1102.0±44.36ns        ? ?/sec    1.12  1231.0±31.75ns        ? ?/sec
or_sliced_24     1.02   247.8±17.45ns        ? ?/sec    1.00    241.8±4.06ns        ? ?/sec